I enjoyed this take on med-chem, and I think he's right:
There are a large set of "don't do this". When they predict failure, you usually shouldn't go there as these rules are moderately reliable.
There is an equally large set of "when you encounter this situation, try this" rules. Their positive predictive power is very very low.
Even the negative rule, the what-to-avoid category, aren't as hard as fast as one would like. There are some pretty unlikely-looking drugs out there (fosfomycin, nitroglycerine, suramin, and see that link above for more). These structures aren't telling you to go out and immediately start imitating them, but what they are telling you is that things that you'd throw away can work.
But those rules are still right more often than the "Here's what to do when . . ." ones, as John Alan Tucker is saying. Every experienced medicinal chemist has a head full of these things - reduce basicity to get out of hERG problems, change the logP for blood-brain-barrier penetration, substitute next to a phenol to slow glucuronidation, switch tetrazole/COOH, make a prodrug, change the salt, and on and on. These work, sometimes, but you have to try them every time before moving on to anything more exotic.
And it's the not-always-right nature of the negative rules, coupled with the not-completely-useless nature of the positive ones, that gives everyone room to argue. Someone has always tried XYZ that worked, while someone else has always tried XYZ when it didn't do a thing. Pretty much any time you try to lay down the law about structures that should or shouldn't be made, you can find arguments on the other side. The rule-of-five type guidelines look rather weak when you think about all the exceptions to them, but they look pretty strong when you compare them to all the other rules that people have tried, and so on.
In the end, all we can do is narrow our options down from an impossible number to a highly improbable number. When (or if) we can do better, medicinal chemistry will change a great deal, but until then. . .